fix: skip None buffers in AutoTP Loading.load_buffer (checkpoint load with nn.InstanceNorm1d)

[SIDDARTHAREDDY8]

Fixes #8514.

Summary
Loading a checkpoint via replace_module/AutoTP into a model containing nn.InstanceNorm1d raised
AttributeError: 'NoneType' object has no attribute 'data' in Loading.load_buffer
(deepspeed/module_inject/auto_tp.py:153). InstanceNorm's running-stat buffers are None until first
use, so module._buffers[name].data.is_meta blows up even though there is nothing to load for them.
The fix skips None buffers in the loop, matching the checkpoint=None walk behavior.

Test notes

• Reproduced the exact AttributeError from the issue on master's load_buffer with an nn.InstanceNorm1d
  module and a saved state dict; post-fix the walk completes and the None buffers are preserved as None.
• Regression-checked that modules with real buffers (e.g. nn.BatchNorm1d) still load checkpoint values
  correctly and that the meta-tensor materialization path is unchanged.
• Full deepspeed package import requires compiled ops, so the check ran the exact shipped function body
  (AST-extracted from the patched file) under torch 2.14.0+cpu. No GPU available here.